中文摘要：

研究了立体视觉的基本原理和空间点的传统三维重建方法；提出基于空间两异面同名光线的公垂线中点的新方法进行空间任意点的三维重建，并验证该意义上的最优解即为满足最小均方误差的最小二乘解；最后在三维重建点的解析解基础上，采用概率描述法定量研究其不确定性，并用基于Mahalanobis距离的超椭球体直观表示其不确定性大小。

英文摘要：

The basic principle of stereo vision and the traditional method to spatial point 3D reconstruction are researched. A new method based on mid-point of common vertical line of two spatial homonymous ray which lie in different spatial plane and they neither parallel nor overlap is proposed to 3D reconstruct any spatial point. And it is validated that the most excellent solution in this meaning is equal to the solution by method of least squares which satisfies the least mean square error. Finally on the base of analytic solution of 3D point which is reconstructed, its uncertainty is researched quantificationally by taking description method with probability, and it is showed intuitionisticly with super ellipsoid founded on Mahalanobis distance.